Evan Kraft American University Dubrovnik, 4 June 2017 Comments on “The Impact of Bank Shocks on Firm-Level Outcomes and Bank Risk-Taking” Evan Kraft American University Dubrovnik, 4 June 2017

What this literature is about Economists have long believed that shocks to banks’ balance sheets affect credit supply Bernanke’s 1983 paper on the Great Depression marks the beginning of this modern literature Idea: A credit crunch is the result of “a significant leftward shift in the supply curve for loans holding constant the safe real interest rate and the quality of potential borrowers” (Bernanke and Lown 1991) Problem: Very hard to disentangle credit supply and demand, especially with aggregate data

Subsequent Advances Use of exogenous shocks that clearly affect supply only Reasonably convincing but not very generalizable Only limited and partial view into aggregate effects Matched bank-firm data Kwaja and Mian 2008, Amiti and Weinstein (2011, 2016) Jimenez, Ongena, Peydro and Saurina (2014) Data at the level of bank loans to particular firms Decompose loan changes into supply and demand shocks Crucial to have multiple bank relationships for each firm to allow identification

Background on the single-bank problem Khwaja and Mian simply assume (fingers crossed!) that the shocks identified using multiple-bank firms are correlated to the shocks to individual bank firms Amor and Nagengast (2016) show for Portugal that firms borrowing from only one bank account for a very small share of total credit. Estimation of supply shocks with all firms and with multiple-bank firms are extremely correlated, so luckily the single bank firms are not a problem for the methodology This paper shows that the same is not true for single-bank firms

This paper’s contribution: including single-bank firms Using Belgian data, Degryse et al show that the characteristics of single-bank firms in Belgium are substantially different than those of multi-bank firms Also, the single-bank firms receive a major share of total loans This suggests the need to keep the single-bank firms in the analysis The solution proposed is aggregation into Industry-Location-Size clusters.

Trade-offs The paper’s approach avoids throwing out over 80% of the firms and over 40% of loans, which makes perfect sense But it risks errors of aggregation in creating the ILS groups The exercises in Table 2 show that the various aggregations produce demand shocks highly correlated to simple firm shocks But it would be nice to know more about How many ILS groups there are How big the variance within these groups is regarding Firm size Output growth Investment ratios

Real impacts: more could be done The paper documents that credit supply shocks do have impacts on real variables: growth, investment and employment However, for some reason there are no summary statistics, sample size in Table 3 showing this. Also, the specification is given without discussion. Would it make sense to include lagged dependent variables? This seems like an important part of the paper, but it is not given too much emphasis

Credit shocks and the crisis The paper is very strong on explaining and exploring the methodology The paper does less to bring out the economic significance of the results A closer analysis of the role of credit demand and supply in the real downturn in 2008-9 would give the paper more of a flavor of policy relevance. Example of what can be done: Amor and Nagengast find 20-40% of investment dynamics in Portugal are due to idiosyncratic bank shocks.

Excellent work, a substantial contribution to an important discussion!